Development of Hybrid Adaptive Neural Fuzzy Inference System-Based Evolutionary Algorithms for Flexural Capacity Prediction in Corroded Steel Reinforced Concrete Beam

Abstract

The damages in reinforced concrete (RC) beams due to reinforcement corrosion is a major problem in the RC industry. Accurate prediction of the residual bearing capacity of RC beams can effectively prevent structural failures or unwanted over-costs of inspections and rehabilitations. This paper proposes a novel machine learning-based prediction framework that combines the adaptive neural fuzzy inference system (ANFIS) with several metaheuristic algorithms for the effective estimation of the flexural strength capacity. Five optimization algorithms are employed for auto-selection of the optimum ANFIS parameters, including differential evolution (DE), genetic algorithm, particle swarm optimization, artificial bee colony, and firefly algorithm (FFA). A comprehensive experimental database of the flexural capacity of corroded steel reinforced concrete beams obtained from the literature, consisting of 177 tests, is used as a case study to evaluate the prediction performance of the proposed hybrid models. The results demonstrate that the proposed hybrid models transcend the previously developed models, while the optimized ANFIS using FFA represents the highest accuracy and strong stability among the proposed models. It is concluded that the proposed framework using ANFIS-FFA can be effectively employed as a useful tool for the accurate estimation of the flexural strength capacity of corroded reinforced concrete beams.